Introducing Explainable AI with ALiX™
Advances in AI are producing autonomous systems that perceive, learn, decide and act on their own. However, the effectiveness of these systems is limited by the machine‘s current inability to explain their decisions and actions to human users. Modal Technology Corporation is introducing ALiX, a new machine learning process that unlocks the ability for the next generation of AI applications to be developed.
ALiX identifies salient data and neural network parameters to reduce complexity, improve knowledge, and explain data relationships previously invisible to human or machine interpretation, so that new physical innovations can be extracted by the user from the AI process.
Training the Machines to Learn Faster
Case studies show the time, money and other resources used to train AI with current industry approaches is rapidly getting out of hand due to the exponential growth in the size and complexity of over‑parameterized neural networks. The capabilities of modern hardware max out when an over‑parameterized neural network reaches ten billion parameters, and there are no obvious advances in current hardware processing technology capability that can keep pace with the exponential growth rate.
Due to the unique explainable AI capabilities of ALiX, Modal Technology Corporation is demonstrating the ability to reverse the exponential growth curve by returning to smaller neural networks that have as‑good or better accuracy than over‑parameterized neural networks. ALiX can find the optimal training solution for the smaller neural network in a single run, and the single run can be parallelized across a very large number of computers in a data center.